Mode Decision using Inter-view Dependencies and Depth Information for Multiview Video Coding

Abstract

Multiview Video Coding (MVC) performs both disparity and motion estimation to exploit inter-view correlations along with the rate distortion optimization to achieve superior coding efficiency, however the encoder load dramatically increases as well. To reduce the stunning computational complexity, many research works focus on fast motion estimation, fast disparity estimation and/or fast mode decision, but are not fully utilized the information supplied by inter-view referencing frames. In this thesis, a fast mode decision algorithm using both inter-view dependencies and depth information is proposed. First, we delicately arrange the minimal candidate modes for each macroblock according to the valued modes corresponding to its inter-view referencing macroblocks, along with the threshold-based early termination to end up the mode decision procedure earlier. Next, we use the result of motion estimation to decide if disparity estimation is required. Third, the search ranges of motion estimation and disparity estimation will be reduced accordingly. Finally, apply the depth information (if available) to refine the search range of disparity estimation and filter out the inappropriate referencing macroblocks as well. The experimental results demonstrate that the proposed algorithm reduces around 80% of the entire encoding time with negligible PSNR drop and bit-rate increasing.